Failure of prospective validation and derivation of a refined clinical decision rule for chest radiography in emergency department patients with chest pain and possible acute coronary syndrome.

OBJECTIVES The authors previously derived a clinical decision rule (CDR) for chest radiography in patients with chest pain and possible acute coronary syndrome (ACS) consisting of the absence of three predictors: history of congestive heart failure, history of smoking, and abnormalities on lung auscultation. The aim of the investigation was to prospectively validate and refine the CDR for chest radiography in an independent patient population. METHODS Patients over 24 years of age with a primary complaint of chest pain and possible ACS were prospectively enrolled from September 2008 to January 2010 at an academic emergency department (ED) with 73,000 annual patient visits. Physicians completed standardized data collection forms before ordering chest radiographs. Two investigators, blinded to the data collection forms, independently classified chest radiographs as "normal,""abnormal not requiring intervention," or "abnormal requiring intervention" (e.g., heart failure, infiltrates), based on review of the radiology report and medical record. Analyses included descriptive statistics, interrater reliability assessment (kappa), and recursive partitioning. RESULTS Of 1,159 visits for possible ACS in which chest radiography was obtained, mean (±SD) age was 60.3 (±15.6) years, and 51% were female. Twenty-four percent had a history of acute myocardial infarction, 10% congestive heart failure, and 11% atrial fibrillation. Sixty-nine (6.0%, 95% confidence interval [CI] = 4.7% to 7.5%) patients had a radiographic abnormality requiring intervention. The kappa statistic for chest radiograph classification was 0.93 (95% CI = 0.88 to 0.97). The previously derived prediction rule (no history of congestive heart failure, no history of smoking, and no abnormalities on lung auscultation) was 78.3% sensitive (95% CI = 67.2% to 86.4%) and 45.1% specific (95% CI = 42.2% to 48.1%) and had a positive predictive value of 8.3% (95% CI = 6.4% to 10.7%) and a negative predictive value of 97.0% (95% CI = 95.2% to 98.2%). Due to suboptimal performance, the rule was refined. The refined rule (no shortness of breath, no history of smoking, no abnormalities on lung auscultation, and age < 55 years) was 100.0% sensitive (95% CI = 93.4% to 100.0%) and 11.5% specific (95% CI = 9.6% to 13.5%) and had a positive predictive value of 6.7% (95% CI = 5.3% to 8.4%) and a negative predictive value of 100.0% (95% CI = 96.3% to 100.0%). CONCLUSIONS Prospective validation of our previously derived CDR for clinically important chest radiographic abnormalities was not successful. Derivation of a refined rule identified all clinically important radiographic abnormalities, but was insufficiently specific. No CDR with adequate sensitivity and specificity could be found.